Statistical Testing Tool Kit Goodness of Fit tests

Statistical Testing Tool. Kit ✦ Goodness of Fit tests ✦ ✦ Useful tookit for experiments (monitoring and validation) Algorithms for binned distributions: Anderson-Darling ✦ Chi-squared ✦ Cramer-von Mises ✦ Ti. Ku ✦ ✦ Algorithms for un-binned distributions: ✦ ✦ ✦ ✦ Anderson-Darling Goodman test Cramer-von Mises Kolmogorov-Smirnov Tiku Girone Watson More info with power of the tests in CHEP 06 presentation ✦ http: //indico. cern. ch/material. Display. py? contrib. Id=305& session. Id=4& material. Id=slides& conf. Id=048 1

Proposal for Statistical Tool. Kit ✦ Package size: ~ 40 files ✦ few 1000 lines of code ✦ Library ~ 300 Kb ✦ ✦ External dependency : ✦ ✦ GSL ROOT interface is just one header file at the moment only for TH 1 objects ✦ Tree interface should be added ? ✦ ✦ Add Statistical Toolkit in a new library and cvs directory will depend on Math. More or GSL (see next slide) ✦ to decide if: ✦ build from a tar file ✦ having the files directly in CVS ✦ ✦ preferable solution 2

Proposal for GSL Some users requested access to all functions of GSL ✦ Build and distribute also GSL ✦ have in ROOT a gsl directory ✦ build GSL from a tar file (as is done in mathmore) ✦ ✦ ✦ expose library and the include file build eventually in ROOT the dictionary for the GSL functions make mathmore dependent of the new lib. Root. GSL ✦ make this changes when the new version of GSL (1. 8) will be released ✦ ✦ will have a fix for Mac. Tel 3